Archived from groups: comp.dcom.lans.ethernet (More info?)
We've got a weird speed problem at work. Here's the setup:
wireless link to our main building
^
|
|
|
|
v
wireless hardware on our building
^
|
|
|
v
router
^
|
|
|
v
switch
^
|
|
v
hub
There are a bunch of computers connected to the switch, and a bunch
connected to the hub. Here's what we see on *most* of our computers:
1. Connections between computers in our building are fast.
2. Uploading data from our building to the main building is fast. (Around
1.2 megabytes/second, which is the expected speed of the wireless link).
3. Downloading data from the main building to ours is slow (30 kbytes/second
is common, with frequent fluctuations up to around 80-100 kbytes/second, and
occasional bursts up to 1 megabyte/second).
I said "most of our computers", because some get fast downloads. I have a
test lab, which has more computers than there are network jacks in the room,
so I brought in an unused DLink home router/switch from home, and am using
the switch portion of it. So, one of the network jacks is connected to one
of the 4 "LAN" ports on the router, and then 3 test computers are hooked to
the others. These all get fast downloads.
Upon experimenting around the building, what we've basically found is this:
1. If a computer is plugged into a network jack, it is slow to download
(except for a single exception, which I'll cover later). It doesn't matter
where that jack connects to on the other end--we've tried moving the other
end from the hub to the switch, or the switch to the hub. It's slow no
matter what.
2. If we take one of the slow computers, and instead of plugging it directly
into the wall jack, we plug it into a cheap switch, and plug the switch into
the wall jack, that computer gets fast downloads.
3. It doesn't seem to have anything to do with the OS. We see this on
Windows, Linux (SuSE and Gentoo), and my PowerBook. It doesn't seem to be a
specific brand of network card in the PCs, either...we've got a variety of
onboard and add-in cards.
The one exception is there is one Windows box that is fast even when plugged
straight into the wall jack. It's specific to that computer--if we unplug
the cable from the back of it and move it over to my PowerBook, for example,
my PowerBook is slow (and then if we stick a switch between the PowerBook
and the wall, the PowerBook is fast).
We also tried using a hub instead of a switch between the computers and the
wall. They stayed slow. (And if we went "computer -> switch -> hub ->
wall", they were fast).
(Oh, and it isn't something freaky about the wiring between the jacks and
the hub or switch, because it was slow when we took my PowerBook into the
wiring closet and plugged it right into the hub with a short cable that is
known to be good).
It would seem that the fact that it only affects downloads from the other
building would indicate a problem with the wireless hardware, or the router.
However, if the problem is on that side of the switch, then it would seem
that it would not be affected by whether or not the computers go through an
extra switch between them and the wall jacks. Conversely, it would seem
that anything that would be affected by that would affect internal traffic
as well as traffic from the other building.
Anyone have any ideas as to what the heck could be going on here?
--
--Tim Smith
We've got a weird speed problem at work. Here's the setup:
wireless link to our main building
^
|
|
|
|
v
wireless hardware on our building
^
|
|
|
v
router
^
|
|
|
v
switch
^
|
|
v
hub
There are a bunch of computers connected to the switch, and a bunch
connected to the hub. Here's what we see on *most* of our computers:
1. Connections between computers in our building are fast.
2. Uploading data from our building to the main building is fast. (Around
1.2 megabytes/second, which is the expected speed of the wireless link).
3. Downloading data from the main building to ours is slow (30 kbytes/second
is common, with frequent fluctuations up to around 80-100 kbytes/second, and
occasional bursts up to 1 megabyte/second).
I said "most of our computers", because some get fast downloads. I have a
test lab, which has more computers than there are network jacks in the room,
so I brought in an unused DLink home router/switch from home, and am using
the switch portion of it. So, one of the network jacks is connected to one
of the 4 "LAN" ports on the router, and then 3 test computers are hooked to
the others. These all get fast downloads.
Upon experimenting around the building, what we've basically found is this:
1. If a computer is plugged into a network jack, it is slow to download
(except for a single exception, which I'll cover later). It doesn't matter
where that jack connects to on the other end--we've tried moving the other
end from the hub to the switch, or the switch to the hub. It's slow no
matter what.
2. If we take one of the slow computers, and instead of plugging it directly
into the wall jack, we plug it into a cheap switch, and plug the switch into
the wall jack, that computer gets fast downloads.
3. It doesn't seem to have anything to do with the OS. We see this on
Windows, Linux (SuSE and Gentoo), and my PowerBook. It doesn't seem to be a
specific brand of network card in the PCs, either...we've got a variety of
onboard and add-in cards.
The one exception is there is one Windows box that is fast even when plugged
straight into the wall jack. It's specific to that computer--if we unplug
the cable from the back of it and move it over to my PowerBook, for example,
my PowerBook is slow (and then if we stick a switch between the PowerBook
and the wall, the PowerBook is fast).
We also tried using a hub instead of a switch between the computers and the
wall. They stayed slow. (And if we went "computer -> switch -> hub ->
wall", they were fast).
(Oh, and it isn't something freaky about the wiring between the jacks and
the hub or switch, because it was slow when we took my PowerBook into the
wiring closet and plugged it right into the hub with a short cable that is
known to be good).
It would seem that the fact that it only affects downloads from the other
building would indicate a problem with the wireless hardware, or the router.
However, if the problem is on that side of the switch, then it would seem
that it would not be affected by whether or not the computers go through an
extra switch between them and the wall jacks. Conversely, it would seem
that anything that would be affected by that would affect internal traffic
as well as traffic from the other building.
Anyone have any ideas as to what the heck could be going on here?
--
--Tim Smith
ccasional bursts up to 1 megabyte/second).
Facebook
Twitter
Instagram